⬆️ bump brainstate from 0.4.0 to 0.4.1

[dependabot]

Bumps brainstate from 0.4.0 to 0.4.1.

Release notes

Sourced from brainstate's releases.

brainstate 0.4.1

A focused patch release that hardens the shared state-aware mapping engine behind vmap / pmap / map (and their module-level *2 variants) against a set of correctness edge cases surfaced by a JAX-expert audit, alongside a routine CI and developer-dependency refresh. No public APIs change.

Bug Fixes

• Read–modify–write states no longer accumulate a spurious axis under mapping: an undeclared state that a mapped function reads and writes in place, and whose shape already matches the mapped axes, is now auto-promoted to a per-lane input and output. Previously each call grew an extra leading axis on the state's value (#203).
• pmap2 now rejects positional argument indices it cannot honor: static_broadcasted_argnums and donate_argnums are no longer silently accepted, because those indices addressed the wrapper's internally bundled arguments rather than the user's. Passing them now raises an explicit error (#203).
• Stale plan cache after state garbage collection: the mapping engine's plan cache is now weakref-backed. When any state captured by a cached plan has been garbage-collected — for example after a module is re-initialized — the plan is rebuilt instead of scattering writes onto orphaned State objects (#203).
• Random sampling inside batched map: drawing random numbers within map(..., batch_size=...) is now supported (#203).
• Consistent replication of non-batched states in the legacy vmap_new_states: NonBatchState / INIT_NO_BATCHING states created inside vmap_new_states are now replicated rather than batched along axis 0, matching the behavior of vmap2_new_states (#203).

Internal Changes

• Consolidated the new-state resolver and the INIT_NO_BATCHING sentinel into the shared _mapping_core module, re-exported from _mapping2 to preserve backward compatibility (#203).
• Documented and hardened the zero-placeholder shape probe and value-dependent control flow, multi-pass (Python-level) side effects, the double init_all_states pass, and the engine's thread-safety guarantees (audit items B4, B7–B10) (#203).
• Merged the standalone composition and nested-leak test suites into the primary _mapping1 / _mapping2 / _mapping_core test modules; the full suite reports 4645 passed, 24 skipped (#203).

CI/CD

• Bumped codecov/codecov-action from v5 to v7 (#199, #202).
• Bumped actions/cache from v4 to v5 (#200).
• Refreshed development dependencies (braintools, mypy) in requirements-dev.txt (#201).

Full Changelog: chaobrain/brainstate@v0.4.0...v0.4.1

Changelog

Sourced from brainstate's changelog.

Version 0.4.1 (2026-06-09)

A focused patch release that hardens the shared state-aware mapping engine behind vmap / pmap / map (and their module-level *2 variants) against a set of correctness edge cases surfaced by a JAX-expert audit, alongside a routine CI and developer-dependency refresh. No public APIs change.

Bug Fixes

• Read–modify–write states no longer accumulate a spurious axis under mapping: an undeclared state that a mapped function reads and writes in place, and whose shape already matches the mapped axes, is now auto-promoted to a per-lane input and output. Previously each call grew an extra leading axis on the state's value (#203).
• pmap2 now rejects positional argument indices it cannot honor: static_broadcasted_argnums and donate_argnums are no longer silently accepted, because those indices addressed the wrapper's internally bundled arguments rather than the user's. Passing them now raises an explicit error (#203).
• Stale plan cache after state garbage collection: the mapping engine's plan cache is now weakref-backed. When any state captured by a cached plan has been garbage-collected — for example after a module is re-initialized — the plan is rebuilt instead of scattering writes onto orphaned State objects (#203).
• Random sampling inside batched map: drawing random numbers within map(..., batch_size=...) is now supported (#203).
• Consistent replication of non-batched states in the legacy vmap_new_states: NonBatchState / INIT_NO_BATCHING states created inside vmap_new_states are now replicated rather than batched along axis 0, matching the behavior of vmap2_new_states (#203).

Internal Changes

• Consolidated the new-state resolver and the INIT_NO_BATCHING sentinel into the shared _mapping_core module, re-exported from _mapping2 to preserve backward compatibility (#203).
• Documented and hardened the zero-placeholder shape probe and value-dependent control flow, multi-pass (Python-level) side effects, the double init_all_states pass, and the engine's thread-safety guarantees (audit items B4, B7–B10) (#203).
• Merged the standalone composition and nested-leak test suites into the primary _mapping1 / _mapping2 / _mapping_core test modules; the full suite reports 4645 passed, 24 skipped (#203).

CI/CD

• Bumped codecov/codecov-action from v5 to v7 (#199, #202).
• Bumped actions/cache from v4 to v5 (#200).
• Refreshed development dependencies (braintools, mypy) in requirements-dev.txt (#201).

Commits

• 9b79925 ci: add codecov.yml with explicit 90% coverage gate (#205)
• d675b5a release: 0.4.1 (#204)
• ff7fccb deps(deps): bump codecov/codecov-action from 6 to 7 (#202)
• 1f015d7 fix(transform): harden state-aware mapping engine (audit B1-B10) (#203)
• fef6dd6 deps(deps-dev): bump the development-dependencies group with 2 updates (#201)
• 7b405b9 deps(deps): bump actions/cache from 4 to 5 (#200)
• e4efc65 deps(deps): bump codecov/codecov-action from 5 to 6 (#199)
• See full diff in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

[dependabot]

Superseded by #94.